module/zfs/zpl_inode.c

   1 /*
   2  * CDDL HEADER START
   3  *
   4  * The contents of this file are subject to the terms of the
   5  * Common Development and Distribution License (the "License").
   6  * You may not use this file except in compliance with the License.
   7  *
   8  * You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE
   9  * or http://www.opensolaris.org/os/licensing.
  10  * See the License for the specific language governing permissions
  11  * and limitations under the License.
  12  *
  13  * When distributing Covered Code, include this CDDL HEADER in each
  14  * file and include the License file at usr/src/OPENSOLARIS.LICENSE.
  15  * If applicable, add the following below this CDDL HEADER, with the
  16  * fields enclosed by brackets "[]" replaced with your own identifying
  17  * information: Portions Copyright [yyyy] [name of copyright owner]
  18  *
  19  * CDDL HEADER END
  20  */
  21 /*
  22  * Copyright (c) 2011, Lawrence Livermore National Security, LLC.
  23  */
  24
  25
  26 #include <sys/zfs_vfsops.h>
  27 #include <sys/zfs_vnops.h>
  28 #include <sys/zfs_znode.h>
  29 #include <sys/vfs.h>
  30 #include <sys/zpl.h>
  31
  32
  33 static struct dentry *
  34 #ifdef HAVE_LOOKUP_NAMEIDATA
  35 zpl_lookup(struct inode *dir, struct dentry *dentry, struct nameidata *nd)
  36 #else
  37 zpl_lookup(struct inode *dir, struct dentry *dentry, unsigned int flags)
  38 #endif
  39 {
  40         cred_t *cr = CRED();
  41         struct inode *ip;
  42         int error;
  43
  44         if (dlen(dentry) > ZFS_MAXNAMELEN)
  45                 return ERR_PTR(-ENAMETOOLONG);
  46
  47         crhold(cr);
  48         error = -zfs_lookup(dir, dname(dentry), &ip, 0, cr, NULL, NULL);
  49         ASSERT3S(error, <=, 0);
  50         crfree(cr);
  51
  52         spin_lock(&dentry->d_lock);
  53         dentry->d_time = jiffies;
  54 #ifndef HAVE_S_D_OP
  55         d_set_d_op(dentry, &zpl_dentry_operations);
  56 #endif /* HAVE_S_D_OP */
  57         spin_unlock(&dentry->d_lock);
  58
  59         if (error) {
  60                 if (error == -ENOENT)
  61                         return d_splice_alias(NULL, dentry);
  62                 else
  63                         return ERR_PTR(error);
  64         }
  65
  66         return d_splice_alias(ip, dentry);
  67 }
  68
  69 void
  70 zpl_vap_init(vattr_t *vap, struct inode *dir, zpl_umode_t mode, cred_t *cr)
  71 {
  72         vap->va_mask = ATTR_MODE;
  73         vap->va_mode = mode;
  74         vap->va_uid = crgetfsuid(cr);
  75
  76         if (dir && dir->i_mode & S_ISGID) {
  77                 vap->va_gid = KGID_TO_SGID(dir->i_gid);
  78                 if (S_ISDIR(mode))
  79                         vap->va_mode |= S_ISGID;
  80         } else {
  81                 vap->va_gid = crgetfsgid(cr);
  82         }
  83 }
  84
  85 static int
  86 #ifdef HAVE_CREATE_NAMEIDATA
  87 zpl_create(struct inode *dir, struct dentry *dentry, zpl_umode_t mode,
  88     struct nameidata *nd)
  89 #else
  90 zpl_create(struct inode *dir, struct dentry *dentry, zpl_umode_t mode,
  91     bool flag)
  92 #endif
  93 {
  94         cred_t *cr = CRED();
  95         struct inode *ip;
  96         vattr_t *vap;
  97         int error;
  98
  99         crhold(cr);
 100         vap = kmem_zalloc(sizeof(vattr_t), KM_SLEEP);
 101         zpl_vap_init(vap, dir, mode, cr);
 102
 103         error = -zfs_create(dir, dname(dentry), vap, 0, mode, &ip, cr, 0, NULL);
 104         if (error == 0) {
 105                 VERIFY0(zpl_xattr_security_init(ip, dir, &dentry->d_name));
 106                 VERIFY0(zpl_init_acl(ip, dir));
 107                 d_instantiate(dentry, ip);
 108         }
 109
 110         kmem_free(vap, sizeof(vattr_t));
 111         crfree(cr);
 112         ASSERT3S(error, <=, 0);
 113
 114         return (error);
 115 }
 116
 117 static int
 118 zpl_mknod(struct inode *dir, struct dentry *dentry, zpl_umode_t mode,
 119     dev_t rdev)
 120 {
 121         cred_t *cr = CRED();
 122         struct inode *ip;
 123         vattr_t *vap;
 124         int error;
 125
 126         /*
 127          * We currently expect Linux to supply rdev=0 for all sockets
 128          * and fifos, but we want to know if this behavior ever changes.
 129          */
 130         if (S_ISSOCK(mode) || S_ISFIFO(mode))
 131                 ASSERT(rdev == 0);
 132
 133         crhold(cr);
 134         vap = kmem_zalloc(sizeof(vattr_t), KM_SLEEP);
 135         zpl_vap_init(vap, dir, mode, cr);
 136         vap->va_rdev = rdev;
 137
 138         error = -zfs_create(dir, dname(dentry), vap, 0, mode, &ip, cr, 0, NULL);
 139         if (error == 0) {
 140                 VERIFY0(zpl_xattr_security_init(ip, dir, &dentry->d_name));
 141                 VERIFY0(zpl_init_acl(ip, dir));
 142                 d_instantiate(dentry, ip);
 143         }
 144
 145         kmem_free(vap, sizeof(vattr_t));
 146         crfree(cr);
 147         ASSERT3S(error, <=, 0);
 148
 149         return (error);
 150 }
 151
 152 static int
 153 zpl_unlink(struct inode *dir, struct dentry *dentry)
 154 {
 155         cred_t *cr = CRED();
 156         int error;
 157
 158         crhold(cr);
 159         error = -zfs_remove(dir, dname(dentry), cr);
 160         crfree(cr);
 161         ASSERT3S(error, <=, 0);
 162
 163         return (error);
 164 }
 165
 166 static int
 167 zpl_mkdir(struct inode *dir, struct dentry *dentry, zpl_umode_t mode)
 168 {
 169         cred_t *cr = CRED();
 170         vattr_t *vap;
 171         struct inode *ip;
 172         int error;
 173
 174         crhold(cr);
 175         vap = kmem_zalloc(sizeof(vattr_t), KM_SLEEP);
 176         zpl_vap_init(vap, dir, mode | S_IFDIR, cr);
 177
 178         error = -zfs_mkdir(dir, dname(dentry), vap, &ip, cr, 0, NULL);
 179         if (error == 0) {
 180                 VERIFY0(zpl_xattr_security_init(ip, dir, &dentry->d_name));
 181                 VERIFY0(zpl_init_acl(ip, dir));
 182                 d_instantiate(dentry, ip);
 183         }
 184
 185         kmem_free(vap, sizeof(vattr_t));
 186         crfree(cr);
 187         ASSERT3S(error, <=, 0);
 188
 189         return (error);
 190 }
 191
 192 static int
 193 zpl_rmdir(struct inode * dir, struct dentry *dentry)
 194 {
 195         cred_t *cr = CRED();
 196         int error;
 197
 198         crhold(cr);
 199         error = -zfs_rmdir(dir, dname(dentry), NULL, cr, 0);
 200         crfree(cr);
 201         ASSERT3S(error, <=, 0);
 202
 203         return (error);
 204 }
 205
 206 static int
 207 zpl_getattr(struct vfsmount *mnt, struct dentry *dentry, struct kstat *stat)
 208 {
 209         boolean_t issnap = ITOZSB(dentry->d_inode)->z_issnap;
 210         int error;
 211
 212         /*
 213          * Ensure MNT_SHRINKABLE is set on snapshots to ensure they are
 214          * unmounted automatically with the parent file system.  This
 215          * is done on the first getattr because it's not easy to get the
 216          * vfsmount structure at mount time.  This call path is explicitly
 217          * marked unlikely to avoid any performance impact.  FWIW, ext4
 218          * resorts to a similar trick for sysadmin convenience.
 219          */
 220         if (unlikely(issnap && !(mnt->mnt_flags & MNT_SHRINKABLE)))
 221                 mnt->mnt_flags |= MNT_SHRINKABLE;
 222
 223         error = -zfs_getattr_fast(dentry->d_inode, stat);
 224         ASSERT3S(error, <=, 0);
 225
 226         return (error);
 227 }
 228
 229 static int
 230 zpl_setattr(struct dentry *dentry, struct iattr *ia)
 231 {
 232         struct inode *ip = dentry->d_inode;
 233         cred_t *cr = CRED();
 234         vattr_t *vap;
 235         int error;
 236
 237         error = inode_change_ok(ip, ia);
 238         if (error)
 239                 return (error);
 240
 241         crhold(cr);
 242         vap = kmem_zalloc(sizeof(vattr_t), KM_SLEEP);
 243         vap->va_mask = ia->ia_valid & ATTR_IATTR_MASK;
 244         vap->va_mode = ia->ia_mode;
 245         vap->va_uid = KUID_TO_SUID(ia->ia_uid);
 246         vap->va_gid = KGID_TO_SGID(ia->ia_gid);
 247         vap->va_size = ia->ia_size;
 248         vap->va_atime = ia->ia_atime;
 249         vap->va_mtime = ia->ia_mtime;
 250         vap->va_ctime = ia->ia_ctime;
 251
 252         error = -zfs_setattr(ip, vap, 0, cr);
 253         if (!error && (ia->ia_valid & ATTR_MODE))
 254                 error = zpl_chmod_acl(ip);
 255
 256         kmem_free(vap, sizeof(vattr_t));
 257         crfree(cr);
 258         ASSERT3S(error, <=, 0);
 259
 260         return (error);
 261 }
 262
 263 static int
 264 zpl_rename(struct inode *sdip, struct dentry *sdentry,
 265     struct inode *tdip, struct dentry *tdentry)
 266 {
 267         cred_t *cr = CRED();
 268         int error;
 269
 270         crhold(cr);
 271         error = -zfs_rename(sdip, dname(sdentry), tdip, dname(tdentry), cr, 0);
 272         crfree(cr);
 273         ASSERT3S(error, <=, 0);
 274
 275         return (error);
 276 }
 277
 278 static int
 279 zpl_symlink(struct inode *dir, struct dentry *dentry, const char *name)
 280 {
 281         cred_t *cr = CRED();
 282         vattr_t *vap;
 283         struct inode *ip;
 284         int error;
 285
 286         crhold(cr);
 287         vap = kmem_zalloc(sizeof(vattr_t), KM_SLEEP);
 288         zpl_vap_init(vap, dir, S_IFLNK | S_IRWXUGO, cr);
 289
 290         error = -zfs_symlink(dir, dname(dentry), vap, (char *)name, &ip, cr, 0);
 291         if (error == 0) {
 292                 VERIFY0(zpl_xattr_security_init(ip, dir, &dentry->d_name));
 293                 d_instantiate(dentry, ip);
 294         }
 295
 296         kmem_free(vap, sizeof(vattr_t));
 297         crfree(cr);
 298         ASSERT3S(error, <=, 0);
 299
 300         return (error);
 301 }
 302
 303 static void *
 304 zpl_follow_link(struct dentry *dentry, struct nameidata *nd)
 305 {
 306         cred_t *cr = CRED();
 307         struct inode *ip = dentry->d_inode;
 308         struct iovec iov;
 309         uio_t uio;
 310         char *link;
 311         int error;
 312
 313         crhold(cr);
 314
 315         iov.iov_len = MAXPATHLEN;
 316         iov.iov_base = link = kmem_zalloc(MAXPATHLEN, KM_SLEEP);
 317
 318         uio.uio_iov = &iov;
 319         uio.uio_iovcnt = 1;
 320         uio.uio_resid = (MAXPATHLEN - 1);
 321         uio.uio_segflg = UIO_SYSSPACE;
 322
 323         error = -zfs_readlink(ip, &uio, cr);
 324         if (error) {
 325                 kmem_free(link, MAXPATHLEN);
 326                 nd_set_link(nd, ERR_PTR(error));
 327         } else {
 328                 nd_set_link(nd, link);
 329         }
 330
 331         crfree(cr);
 332         return (NULL);
 333 }
 334
 335 static void
 336 zpl_put_link(struct dentry *dentry, struct nameidata *nd, void *ptr)
 337 {
 338         const char *link = nd_get_link(nd);
 339
 340         if (!IS_ERR(link))
 341                 kmem_free(link, MAXPATHLEN);
 342 }
 343
 344 static int
 345 zpl_link(struct dentry *old_dentry, struct inode *dir, struct dentry *dentry)
 346 {
 347         cred_t *cr = CRED();
 348         struct inode *ip = old_dentry->d_inode;
 349         int error;
 350
 351         if (ip->i_nlink >= ZFS_LINK_MAX)
 352                 return -EMLINK;
 353
 354         crhold(cr);
 355         ip->i_ctime = CURRENT_TIME_SEC;
 356         igrab(ip); /* Use ihold() if available */
 357
 358         error = -zfs_link(dir, ip, dname(dentry), cr);
 359         if (error) {
 360                 iput(ip);
 361                 goto out;
 362         }
 363
 364         d_instantiate(dentry, ip);
 365 out:
 366         crfree(cr);
 367         ASSERT3S(error, <=, 0);
 368
 369         return (error);
 370 }
 371
 372 #ifdef HAVE_INODE_TRUNCATE_RANGE
 373 static void
 374 zpl_truncate_range(struct inode* ip, loff_t start, loff_t end)
 375 {
 376         cred_t *cr = CRED();
 377         flock64_t bf;
 378
 379         ASSERT3S(start, <=, end);
 380
 381         /*
 382          * zfs_freesp() will interpret (len == 0) as meaning "truncate until
 383          * the end of the file". We don't want that.
 384          */
 385         if (start == end)
 386                 return;
 387
 388         crhold(cr);
 389
 390         bf.l_type = F_WRLCK;
 391         bf.l_whence = 0;
 392         bf.l_start = start;
 393         bf.l_len = end - start;
 394         bf.l_pid = 0;
 395         zfs_space(ip, F_FREESP, &bf, FWRITE, start, cr);
 396
 397         crfree(cr);
 398 }
 399 #endif /* HAVE_INODE_TRUNCATE_RANGE */
 400
 401 #ifdef HAVE_INODE_FALLOCATE
 402 static long
 403 zpl_fallocate(struct inode *ip, int mode, loff_t offset, loff_t len)
 404 {
 405         return zpl_fallocate_common(ip, mode, offset, len);
 406 }
 407 #endif /* HAVE_INODE_FALLOCATE */
 408
 409 static int
 410 #ifdef HAVE_D_REVALIDATE_NAMEIDATA
 411 zpl_revalidate(struct dentry *dentry, struct nameidata *nd)
 412 {
 413         unsigned int flags = (nd ? nd->flags : 0);
 414 #else
 415 zpl_revalidate(struct dentry *dentry, unsigned int flags)
 416 {
 417 #endif /* HAVE_D_REVALIDATE_NAMEIDATA */
 418         zfs_sb_t *zsb = dentry->d_sb->s_fs_info;
 419         int error;
 420
 421         if (flags & LOOKUP_RCU)
 422                 return (-ECHILD);
 423
 424         /*
 425          * After a rollback negative dentries created before the rollback
 426          * time must be invalidated.  Otherwise they can obscure files which
 427          * are only present in the rolled back dataset.
 428          */
 429         if (dentry->d_inode == NULL) {
 430                 spin_lock(&dentry->d_lock);
 431                 error = time_before(dentry->d_time, zsb->z_rollback_time);
 432                 spin_unlock(&dentry->d_lock);
 433
 434                 if (error)
 435                         return (0);
 436         }
 437
 438         /*
 439          * The dentry may reference a stale inode if a mounted file system
 440          * was rolled back to a point in time where the object didn't exist.
 441          */
 442         if (dentry->d_inode && ITOZ(dentry->d_inode)->z_is_stale)
 443                 return (0);
 444
 445         return (1);
 446 }
 447
 448 const struct inode_operations zpl_inode_operations = {
 449         .create         = zpl_create,
 450         .link           = zpl_link,
 451         .unlink         = zpl_unlink,
 452         .symlink        = zpl_symlink,
 453         .mkdir          = zpl_mkdir,
 454         .rmdir          = zpl_rmdir,
 455         .mknod          = zpl_mknod,
 456         .rename         = zpl_rename,
 457         .setattr        = zpl_setattr,
 458         .getattr        = zpl_getattr,
 459         .setxattr       = generic_setxattr,
 460         .getxattr       = generic_getxattr,
 461         .removexattr    = generic_removexattr,
 462         .listxattr      = zpl_xattr_list,
 463 #ifdef HAVE_INODE_TRUNCATE_RANGE
 464         .truncate_range = zpl_truncate_range,
 465 #endif /* HAVE_INODE_TRUNCATE_RANGE */
 466 #ifdef HAVE_INODE_FALLOCATE
 467         .fallocate      = zpl_fallocate,
 468 #endif /* HAVE_INODE_FALLOCATE */
 469 #if defined(CONFIG_FS_POSIX_ACL)
 470 #if defined(HAVE_GET_ACL)
 471         .get_acl        = zpl_get_acl,
 472 #elif defined(HAVE_CHECK_ACL)
 473         .check_acl      = zpl_check_acl,
 474 #elif defined(HAVE_PERMISSION)
 475         .permission     = zpl_permission,
 476 #endif /* HAVE_GET_ACL | HAVE_CHECK_ACL | HAVE_PERMISSION */
 477 #endif /* CONFIG_FS_POSIX_ACL */
 478 };
 479
 480 const struct inode_operations zpl_dir_inode_operations = {
 481         .create         = zpl_create,
 482         .lookup         = zpl_lookup,
 483         .link           = zpl_link,
 484         .unlink         = zpl_unlink,
 485         .symlink        = zpl_symlink,
 486         .mkdir          = zpl_mkdir,
 487         .rmdir          = zpl_rmdir,
 488         .mknod          = zpl_mknod,
 489         .rename         = zpl_rename,
 490         .setattr        = zpl_setattr,
 491         .getattr        = zpl_getattr,
 492         .setxattr       = generic_setxattr,
 493         .getxattr       = generic_getxattr,
 494         .removexattr    = generic_removexattr,
 495         .listxattr      = zpl_xattr_list,
 496 #if defined(CONFIG_FS_POSIX_ACL)
 497 #if defined(HAVE_GET_ACL)
 498         .get_acl        = zpl_get_acl,
 499 #elif defined(HAVE_CHECK_ACL)
 500         .check_acl      = zpl_check_acl,
 501 #elif defined(HAVE_PERMISSION)
 502         .permission     = zpl_permission,
 503 #endif /* HAVE_GET_ACL | HAVE_CHECK_ACL | HAVE_PERMISSION */
 504 #endif /* CONFIG_FS_POSIX_ACL */
 505 };
 506
 507 const struct inode_operations zpl_symlink_inode_operations = {
 508         .readlink       = generic_readlink,
 509         .follow_link    = zpl_follow_link,
 510         .put_link       = zpl_put_link,
 511         .setattr        = zpl_setattr,
 512         .getattr        = zpl_getattr,
 513         .setxattr       = generic_setxattr,
 514         .getxattr       = generic_getxattr,
 515         .removexattr    = generic_removexattr,
 516         .listxattr      = zpl_xattr_list,
 517 };
 518
 519 const struct inode_operations zpl_special_inode_operations = {
 520         .setattr        = zpl_setattr,
 521         .getattr        = zpl_getattr,
 522         .setxattr       = generic_setxattr,
 523         .getxattr       = generic_getxattr,
 524         .removexattr    = generic_removexattr,
 525         .listxattr      = zpl_xattr_list,
 526 #if defined(CONFIG_FS_POSIX_ACL)
 527 #if defined(HAVE_GET_ACL)
 528         .get_acl        = zpl_get_acl,
 529 #elif defined(HAVE_CHECK_ACL)
 530         .check_acl      = zpl_check_acl,
 531 #elif defined(HAVE_PERMISSION)
 532         .permission     = zpl_permission,
 533 #endif /* HAVE_GET_ACL | HAVE_CHECK_ACL | HAVE_PERMISSION */
 534 #endif /* CONFIG_FS_POSIX_ACL */
 535 };
 536
 537 dentry_operations_t zpl_dentry_operations = {
 538         .d_revalidate   = zpl_revalidate,
 539 };